Delayed coastal inundations caused by ocean dynamics post-Hurricane Matthew

The study addresses the persistent abnormal water levels (PHAWL) observed after hurricanes and their contribution to delayed and nuisance flooding along the U.S. Southeast Coast (USSC). While storm surge, waves, precipitation, and river discharge are well-studied during hurricanes, the post-hurricane oceanic adjustment that sustains elevated sea levels has not been thoroughly examined. Hurricane Matthew (2016) exhibited significant Non-Tidal Residual Anomalies (NTRAs) of 35–50 cm lasting days to weeks along eight NOAA tide gauges in the USSC, indicating long-lasting flood risk. The research question is to quantify the spatiotemporal extent of PHAWL and identify the mechanisms—particularly the roles of atmospheric forcing versus oceanic processes such as Gulf Stream (GS) variability and coastally trapped waves (CTWs)—that sustain elevated sea levels after hurricane passage. The purpose is to improve prediction and planning for coastal protection by including post-hurricane ocean dynamics in total water level assessments.

Multiple oceanic processes modulate USSC coastal sea levels. Thermosteric effects associated with the Florida Current/Gulf Stream (GS) and Atlantic Meridional Overturning Circulation (AMOC) have been linked to interannual to multidecadal sea level variability and coastal flood frequency changes. Hurricanes can significantly weaken the GS (observed in events such as Bill 2009, Joaquin 2015, Dorian 2019, and Matthew 2016), affecting coastal sea level through geostrophic balance. High correlations between GS variability and coastal sea level have been documented, indicating fast barotropic responses. Basin-scale Rossby waves modulate longer-term sea level changes but act on slower timescales. Coastally trapped waves (CTWs) can propagate along the USSC and modulate coastal sea levels on subseasonal timescales; recent work shows southward propagation from Cape Hatteras with amplitudes up to ~15 cm. Gaps remain in understanding high-frequency post-hurricane dynamics: Rossby waves are too slow, thermosteric effects are unlikely immediately post-storm due to cooling, and prior GS–sea level studies often relied on limited spatial observations or monthly means that miss short-term variability. Tide-gauge-only analyses confound multiple forcings. High-resolution, 3-D baroclinic coastal models are needed to resolve the interacting drivers but face challenges such as resolution and mixing errors.

A 3-D, high-resolution, baroclinic coastal ocean model (SCHISM) with an unstructured grid was configured to simulate the US East Coast and Gulf of Mexico (98°W–60°W, 8°N–46°N). Horizontal resolution ranges from ~6 km in the open ocean to ~5 m in inland regions; the highest resolution (~5 m) was applied to the south Georgia coast to resolve complex wetlands and urban areas. Bathymetry/topobathymetry used GEBCO 2022 in the open ocean and CUDEM for the Georgia coast. A polymorphic vertical grid allowed seamless transitions among 3D/2D configurations depending on depth. Surface forcing combined ECMWF and HRRR products. Initial and boundary conditions used CMEMS (temperature, salinity, velocity) and AVISO (sea surface height). Tides were represented using FES2014 (K1, K2, M2, N2, O1, P1, Q1, S2) on the open boundary and tidal potentials in momentum equations to simulate total water levels and inundation depths above ground. The simulation period was 2016-08-01 to 2016-10-31, focusing analyses on 2016-09-28 to 2016-10-22 (Hurricane Matthew and post-hurricane). Model validation used NOAA tide gauges, submarine cable Florida Current data, Argo, AVISO, and JPL-GHRSST; skill assessments are shown in Supplementary Figs. 2–8. Sensitivity experiments separated drivers of PHAWL into atmospheric forcing (ATF) and oceanic adjustment (OCADJ). CTRL included all forcings (precipitation, wind, air pressure, heat flux, oceanic forcing) with tides removed to isolate non-tidal residuals. OCADJ excluded atmospheric forcing, leaving ocean dynamics active; ATF was computed as CTRL minus OCADJ (acknowledging minor nonlinearity). Spatial and temporal diagnostics included Hovmöller diagrams along the USSC, maps of maximum NTRAs, city-scale inundation modeling for south Georgia, decomposition of coastal vs shelf-break signals, and correlations between GS speed (path-averaged current along the GS) and water levels on the continental shelf (particularly the 200 m isobath). CTW propagation was examined using snapshots of SSH anomalies post-storm.

• PHAWL magnitude and duration: After Matthew’s departure (from ~2016-10-10), coastal PHAWLs of 20–54 cm impacted the entire South Atlantic Bight (SAB) through ~2016-10-14, with continued fluctuations thereafter. Maximum post-hurricane anomalies across the shelf reached up to 58 cm relative to pre-storm means; highest peaks (>50 cm) were in south Georgia and Florida.
• Inundation impacts: City-scale modeling on the south Georgia coast showed PHAWL increased inundation depths by up to 30 cm in wetlands and residential areas, implying delayed, nuisance/sunny-day flooding risks post-storm.
• Driver attribution: ATF produced time-varying coastal fluctuations, peaking at ~25 cm on 2016-10-13 (about 42% of PHAWL), then oscillating roughly 1–15 cm. OCADJ peaked early at ~45 cm on 2016-10-11 (>90% of PHAWL), and maintained elevated shelf-scale water levels of ~33–38 cm for ~10 days, setting the mean component and duration.
• Spatial structure: Two main signals were identified—coastal (primarily ATF-driven during certain times) and offshore signals entering from the shelf break (OCADJ-driven). Offshore water levels along the shelf break increased to ~52 cm and propagated shoreward, raising coastal levels by ~25–43 cm depending on location (largest in Florida).
• CTWs: Fast alongshore propagation of positive SSH anomalies from north to south along the SAB coast occurred immediately after storm exit, consistent with CTWs with speeds on the order of ~12 m s−1, producing initial spikes in OCADJ-induced water levels on 1–3 day timescales.
• Gulf Stream linkage: The GS weakened by up to ~50% after Matthew due to mixing and adverse winds, with strong anticorrelation between GS speed and shelf water levels. Correlations between GS speed and shelf-break water levels were high (R ≈ −0.92 with low-pass filter; R ≈ −0.75 without). Florida’s proximity to the shelf break yielded stronger negative correlations (−0.85 to −0.47) than farther north (−0.45 to −0.17), indicating greater vulnerability in Florida.
• Overall mechanism: CTWs set the initial peak and rapid alongshore modulation (1–3 days), while GS-driven cross-shelf signals sustain elevated shelf water levels over >10 days, determining the mean and persistence of PHAWL.

The findings resolve the mechanisms behind persistent post-hurricane elevated sea levels along the USSC, demonstrating that ocean dynamics, rather than atmospheric forcing alone, control the magnitude and duration of PHAWL. The model–observation framework shows that CTWs provide rapid alongshore propagation right after storm passage, explaining the timing and spikes of coastal anomalies, while weakening of the Gulf Stream generates sustained cross-shelf pressure gradients that elevate water levels across the continental shelf. This integrated view addresses the research question by quantifying the distinct roles and timescales of ATF and OCADJ and explaining why delayed, widespread inundation can occur even without direct hurricane landfall at a given location. The results highlight the importance of including oceanic adjustment processes in forecasting and risk assessment, as their large spatial reach implies hazards beyond the storm’s immediate impact zone.

This study introduces PHAWL as a critical component of post-hurricane coastal hazards and quantifies its drivers using a validated, high-resolution 3-D coastal model. It shows that: (1) PHAWLs of up to ~58 cm persisted for weeks after Hurricane Matthew, increasing inundation depths by up to ~30 cm in coastal communities; (2) atmospheric forcing mainly modulates short-term fluctuations, while oceanic adjustment—via CTWs (1–3 days) and GS-driven cross-shelf signals (>10 days)—controls the mean level and duration across the SAB. These insights call for operational modeling frameworks that jointly consider atmospheric and oceanic adjustments to predict total water levels and inundation during and after hurricanes. Future work should employ ensemble experiments spanning diverse hurricane tracks and intensities to assess how GS responses and CTW characteristics generalize, improving probabilistic risk assessments under climate change scenarios with stronger and more frequent storms.

• Case specificity: Analyses focus on a single event (Hurricane Matthew, 2016); generalization to other storms requires ensemble studies with varying tracks and intensities.
• Decomposition nonlinearity: ATF was estimated as CTRL minus OCADJ, which may include minor nonlinear interaction residuals.
• Observational isolation: While CTWs are evident in the model, isolating them unambiguously from tide-gauge data is challenging due to overlapping forcings (tides, winds, pressure).
• Spatial dependence: The strength of GS–sea level coupling varies with proximity to the shelf break, implying geographic variability in applicability and vulnerability.